Since the parking resources become more and more intense, how to make effective management of parking spaces has become a topic for the intelligent transportation system. Compared with the infrared detection and the sensing coil detection, the video analysis method has the advantages of saving hardware cost, easy installation and extending other functions and so on. The algorithms of analyzing location of parking spaces in video are studied, which combine detecting license plate with detecting parking spaces. The boundary coordinates and central coordinates of the license plate region are detected to determine the occupancy of the parking space. The method of detecting the gray level of the parking space is used to determine the occupancy of the vehicle when the vehicle is sheltered by pedestrian. According to the current occupancy status of all the parking spaces, a shortest path method is designed to guide the parking when a vehicle enters the parking space. The experiment shows that recognition rate of the parking space detection system designed in this paper is beyond 90% and the system has a certain degree of stability and a strong practicality.